DeepX (CEO Lokwon Kim), an AI semiconductor company specializing in low-power on-device AI, announced on the 18th that it has signed the Paddle Paddle Technology Ecosystem Partnership.
As the official ecosystem partner based on Baidu's open source deep learning framework paddle paddle, DeepX will collaborate in a variety of industrial AI projects, co-develop products based on technology compatibility, and promote global customer promotion activities.
Baidu's PaddlePaddle is China's first open source deep learning framework, a core technology platform for creating and running AI models. The name stands for “Parallel Distributed Deep Learning” and is designed to quickly process large amounts of data and operate efficiently on a variety of devices.
Also Read: An Aithority Interview with Suzanne Livingston, Vice President of IBM Watsonx Orchestrated Agent Domain
It is used in a wide range of industries including smart cities, autonomous driving, image recognition, voice processing, and more, and is considered China's most widely used AI development tool, and is adopted by thousands of companies and institutions, as well as Baidu's own internal products. In particular, paddle paddles are located as a “full stack AI platform” that supports the entire process from AI model development to deployment.
Before the contract, the DX-M1, a mass-produced chip from DEEPX, achieved extremely high FPS (frames per second) and power efficiency (FPS/W) especially in industrial PCS based on limited battery environments, and was demonstrated in real time with Baidu's 5th generation PP-OCR and VLM-based models. Baidu is keen to further aggressive collaboration in the future.
Cooperation will extend not only to OCR, but also to drones and robots. Baidu PaddlePaddle and Ernie teams will compile current AI models of drones and VLM+ obstacle avoidance models for deployment to robots on the DX-M1 and commercialize them for real-world robotics and drone projects deployment.
Meanwhile, DEEPX transforms and compiles 10 representative OpenVino-based AI models that are widely used in the industry for compatibility with DX-M1, and these models are shared with Baidu's global partners in PaddlePaddle ecosystem. This is expected to make it easier for a variety of customers to adopt and scale AI products with Deepx NPUs.
Deepx is also developing a V-NPU, a Vision AI-only NPU card that can replace GPUs, and completed the initial demo in mass production, scheduled to begin in September. DeepX and Baidu's cooperation on next-generation product development is also being discussed. DEEPX will use Samsung Foundry's next-generation AI semiconductor “DX-M2” based on the 2NM process to verify a prototype that runs Baidu's large AI model “Ernie-4.5-VL-28B-A3B” (MOE-based).
Additionally, Baidu Paddlepaddle and Ernie Team will propose jointly exhibiting Deepx at the 2025 Deep Shenzhen (International) Artificial General Information Conference held at Deep Shenzhen at the end of August, and will also provide local demonstrations and co-marketing to customers around the world.
Baidu PaddlePaddle and the Ernie team said: “As a partner with superior low-power AI semiconductor technology, DEEPX will add powerful edge computing capabilities to the Baidu AI technology ecosystem, particularly the DX-M1. It accelerates ecosystem expansion and makes it easier for more companies to implement AI in real-world industries.”
“The partnership with Baidu represents a key turning point in the rapid spread of DeepX's AI semiconductor technology to global industrial sites,” said Lokwon Kim, CEO of DeepX. “We will strengthen our competitiveness in the global AI market through practical commercialization based on paddle paddles and the Ernie ecosystem and various industrial collaborations.”
Also Read: c-gen.ai impacts AI workloads from stealth to end-infrastructure limits
[To share your insights with us as part of editorial or sponsored content, please write to psen@itechseries.com]
